dodox/reppo
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

README fixes and polish

Browse Source
This commit is contained in:
ys1087@partner.kit.edu 2025-08-14 01:14:35 +02:00
parent ecdffce579
commit 8411abba7b
1 changed files with 5 additions and 17 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

22
README.md
View File

@ -16,6 +16,8 @@ Our repo provides you with the core algorithm and the following features:
### Installation on HoReKa ### Installation on HoReKa
Original Repo recommends ´uv´, but I prefer vanilla python and that seems to work...
1. **Clone the repository and navigate to it:** 1. **Clone the repository and navigate to it:**
```bash ```bash
git clone <repository-url> git clone <repository-url>
@ -42,22 +44,15 @@ The repository includes pre-configured SLURM scripts with wandb integration:
#### Quick Start #### Quick Start
```bash ```bash
# Submit a ManiSkill job
./submit_job.sh maniskill PickCube-v1 mjx_dmc_medium_data
# Submit a Brax job
./submit_job.sh brax ant mjx_dmc_small_data ./submit_job.sh brax ant mjx_dmc_small_data
``` ```
#### Manual Job Submission #### Manual Job Submission
```bash ```bash
# Submit ManiSkill experiments # Submit Brax experiments
sbatch slurm/run_reppo_maniskill.sh
# Submit Brax experiments (completed)
sbatch slurm/run_reppo_brax.sh sbatch slurm/run_reppo_brax.sh
# Submit DMC experiments (new) # Submit DMC experiments
python submit_dmc_experiments.py --seeds 3 python submit_dmc_experiments.py --seeds 3
# With custom environment # With custom environment
@ -67,7 +62,7 @@ ENV_NAME=PlaceApple-v1 EXPERIMENT_TYPE=mjx_dmc_large_data sbatch slurm/run_reppo
#### Supported Environments #### Supported Environments
**ManiSkill environments:** **ManiSkill environments:**
- `PickCube-v1`, `PlaceApple-v1`, `StackCube-v1`, `PegInsertionSide-v1` - `PickCube-v1`, `PlaceApple-v1`, `StackCube-v1`, `PegInsertionSide-v1`, ...
**Brax environments:** **Brax environments:**
- `ant`, `cheetah`, `hopper`, `walker2d`, `humanoid` - `ant`, `cheetah`, `hopper`, `walker2d`, `humanoid`
@ -87,11 +82,6 @@ tail -f logs/reppo_maniskill_<job_id>.out
tail -f logs/reppo_brax_<job_id>.out tail -f logs/reppo_brax_<job_id>.out
``` ```
All experiments automatically log to wandb with your configured credentials. Results will appear in projects:
- `reppo_brax_production` (completed)
- `reppo_dmc_production` (in progress)
- `reppo_maniskill` (pending)
#### Critical Issues in Official Repository #### Critical Issues in Official Repository
⚠️ **The official REPPO repository is not runnable due to a series of fatal bugs.** These issues were discovered and fixed during HoReKa cluster deployment: ⚠️ **The official REPPO repository is not runnable due to a series of fatal bugs.** These issues were discovered and fixed during HoReKa cluster deployment:
@ -128,8 +118,6 @@ All experiments automatically log to wandb with your configured credentials. Res
- **Root cause**: BraxGymnaxWrapper wasn't properly vectorized for multi-environment operations - **Root cause**: BraxGymnaxWrapper wasn't properly vectorized for multi-environment operations
- **Fix applied**: Added proper vectorization support to `reset()` and `step()` methods using `jax.vmap` for handling both single and batched operations - **Fix applied**: Added proper vectorization support to `reset()` and `step()` methods using `jax.vmap` for handling both single and batched operations
**Summary**: Fixed 6 critical bugs that prevented the original repository from running. The algorithm now successfully runs with 256 parallel environments and proper wandb integration, achieving strong learning performance (episode returns improving from ~-100 to ~400+ in ant environment).
--- ---
## Original README ## Original README